Researchers at Edith Cowan University (ECU) are pioneering a novel computer tracking technology that analyzes camera footage to determine whether a driver is impaired by alcohol before they even start driving. This advancement could enhance road safety by preventing impaired drivers from getting behind the wheel.
The Research Process
In collaboration with Mix by Powerfleet, the ECU team conducted controlled experiments where participants with varying levels of alcohol intoxication—sober, low intoxication, and severely intoxicated—were recorded while driving in a simulator. Using standard RGB (red, green, blue) video footage of the drivers’ faces, researchers developed a machine learning system to assess impairment based on facial cues, gaze direction, and head position.
Accuracy and Potential Applications
The system demonstrated an overall accuracy of 75% in classifying drivers’ intoxication levels. Ph.D. student Ensiyeh Keshtkaran emphasized that this technology could be integrated into vehicles equipped with driver monitoring systems or even smartphones, allowing for early detection of intoxicated drivers. Unlike traditional methods that rely on observing driving behaviors, this approach aims to identify impairment before the vehicle is in motion.
Technological Innovation
This research marks a significant breakthrough as it is the first to utilize a standard RGB camera to detect alcohol impairment through facial indicators. Dr. Syed Zulqarnain Gilani, a senior lecturer at ECU, noted that if low-resolution video proves sufficient, this technology could be deployed in roadside surveillance cameras or used by law enforcement to combat drunk driving more effectively.
Broader Implications for Road Safety
Drunk driving accounts for approximately 30% of fatal crashes in Australia, with one in five drivers killed on the roads having a blood alcohol concentration (BAC) of 0.05 or higher. Current detection methods, mainly random breath tests, do not fully address the issue of impaired driving.
Keshtkaran pointed out that existing approaches primarily analyze driving behaviors, which can only be assessed once a driver is already in motion, highlighting a critical gap in proactive safety measures. By leveraging computer vision to detect signs of intoxication based on facial and behavioral changes, this new system aims for faster intervention, enhancing public safety.
Future Directions
As research progresses, the dataset generated from this study—which includes 3D and infrared videos of drivers, rearview footage, and detailed driving simulation logs—will serve as a valuable resource for further exploration in the field. The next steps will focus on refining the technology’s image resolution requirements, potentially paving the way for broader applications in traffic monitoring and enforcement.
In summary, ECU’s innovative approach to detecting drunk drivers could transform road safety measures, offering a proactive solution to a persistent problem.